Lennex Zinyando

I accidentally built a local AI inference engine. Never intended to. Just wanted to test some models. Now it has transcription, TTS, voice cloning—all running locally. Writing custom Metal kernels to make it faster. github.com/izwi-ai/izwi

@trendsdotfun This would be awesome. How would it work?

@zinyando I wonder what other features might be added later? We'd like to invite you to share it at our hackathon event. x.com/trendsdotfun/s…

I accidentally built a local AI inference engine. Never intended to. Just wanted to test some models. Now it has transcription, TTS, voice cloning—all running locally. Writing custom Metal kernels to make it faster. github.com/izwi-ai/izwi

@shaun_kudzi @IzwiAI Not yet, but African language models are definitely coming to Izwi in the near future. Training/fine tuning great models requires money for GPUs, I need to have a bit more capital to do it correctly.

@zinyando @IzwiAI @zinyando this is fascinating. Have you tried the google/WaxalNLP dataset (Diack et al., 2026) Configuration: sna_asr (Shona language subset) from Hugging Face.

200+ stars on GitHub for @IzwiAI. This project started because I wanted to try Qwen-TTS/Qwen-ASR and couldn't find an easy way. So I built it myself. Started by accident. Feels like the right move. github.com/izwi-ai/izwi

Africa left the chat...

200+ GitHub stars 🚀

@m4zas24 I'm working on shipping a stable release first with better performance, then I can start making demo videos. Lots of work to do, but videos are on the cards

@zinyando @zinyando Awesome do you mind showing us a demo or video of how the audio sounds/works etc

@villainmonkey Yeah, my DM is open.

@zinyando Dm?

@diginotive1 You can send me a message, my DMs are open

@zinyando Hello Lennex, kindly inbox

@paulabartabajo_ @liquidai You can use @liquidai audio models on @IzwiAI running locally. It exposes an OpenAI compatible server so you have build your local desktop apps on top of it github.com/izwi-ai/izwi

Advice for AI engineers 💡 A small Audio Language Model transcribes audio in real-time on your device, without internet connection or cloud costs. For example, LFM2.5-Audio-1.5B by @liquidai ↓

@mrstephendeng I’m building @IzwiAI, working on making on-device AI actually useful. Starting with voice workflows such as TTS, transcription, speaker diarization, etc

I'm seeing African founders building some really, really interesting things these days, going deep into what is structurally changing with the use of AI. I mean everything from semiconductors to robotics to oil & gas to pharma to fish. For the first time, our pipeline isn't absolutely dominated by fintech and its happened faster than I thought it would. It's also increasingly clear that the bottleneck on "venture backability" on the continent is shifting from product to customer discovery & sales. Given what's possible, there's so much more market to take, but it's also never been easier to spin up quality software and hardware to tackle exciting problems. However, the very best teams we come across can sell convincingly and early to 'lighthouse' customers and permeate their workflows before the crowd catches up. They can then validate the shape and size of that demand, tailor and compound their intelligence using proprietary data, and go out to lock down a fundraising advantage. Flywheel spins.

@TheDumbTechGuy This is only for LLM inference, not the actual model training. I'd like to fine tune models for TTS and ASR for African languages, but unfortunately, I'd need money for GPUs for that.

@zinyando How are you training it?

"We have vendor assessments." → "Our AI never touches the cloud." "We have DPIA underway" → "GDPR-compliant by design." The difference between a 3-month sales cycle and a 12-month compliance review. izwiai.com/blog/gdpr-by-d…

@TheObiLeonard I’m building @IzwiAI, on-device voice AI engine izwiai.com

I’ll be sponsoring 3 indie builders with the “Claude Pro” plan and 2 others with a “Claude Max” plan for up to three months in credit. You just have to show you can build really cool shit and can build it fast.

AI in the Boardroom: Why the C-Suite Is Banning Cloud AI izwiai.com/blog/ai-in-the…

Your M&A strategy lives in vendor logs. Your SEC investigation discussion sits on someone else's servers. That's not a risk. It's a certainty. On-device transcription keeps your boardroom in the room. izwiai.com/blog/ai-in-the…

@OmriBuilds Practical On-device AI izwiai.com

Pitch your startup in 3 words

@KaiXCreator I'm working on @IzwiAI, On-device AI for voice workflows, transcriptions, speaker diarization, TTS, voice cloning, and design izwiai.com

What are you building this week? Drop your project URL Let’s drive some traffic

@samteic I'm working on @IzwiAI, On-device AI for voice workflows, transcriptions, speaker diarization, TTS, voice cloning, and design izwiai.com

If you’re building in public, reply with what you’re working on. I’m looking for more builders to follow.

@StevBuilds I'm working on @IzwiAI, On-device AI for voice workflows, transcriptions, speaker diarization, TTS, voice cloning, and design izwiai.com

Distribution exercise🚀 Can you explain your product in one sentence? Prove it!👇